Puget Systems

We've built benchmarks for Premiere, After Effects, Photoshop, DaVinci Resolve, and more. Yesterday we launched one that's been a long time coming: Puget Bench for Unreal Engine.

The challenge with benchmarking UE is that no two users are doing the same thing. A game dev building an open world stresses hardware completely differently than a virtual production studio rendering cinematic content in real time even though they're both using the same engine.

So we built around real workflows:
🔧 Shader compilation
🖥️ Viewport performance
🎬 Movie Render Queue output
📦 Asset processing
💡 Lighting builds

The goal isn't a flashy score; it's data that tells you exactly what to upgrade to improve your specific workflow.

Read the full blog post and join the beta program below.
And if you're heading to Unreal Fest in Chicago next week, come find us! We'll be showing this off live!

We're heading to Unreal Fest in Chicago next week (June 16–18), and before we get on the floor we wanted to share what we've actually learned from years of benchmarking and building workstations for Unreal Engine 5.

Because the answer to "What does a good UE5 workstation need?" is: It depends on your role. And that distinction matters a lot more than most spec sheets let on.

For artists and environment creators, the GPU is the most important call.

Modern UE5 workflows are VRAM-hungry. High-resolution textures, Nanite-dense scenes, hardware ray tracing, and Lumen lighting all lean heavily on the GPU. We're recommending the NVIDIA GeForce RTX 5080 as a strong all-around option, with the RTX 5090 stepping in when you're pushing VR content or particularly dense scenes that need every bit of that 32GB buffer. CPU requirements are more modest in this role. Intel Core Ultra performs well here for editor responsiveness and occasional shader builds, without overbuying on platform cost.

For programmers and technical artists, the CPU is the bottleneck.

Unreal Engine and Visual Studio shader compilation scales extremely well with core count and for developers who compile frequently throughout the day, that time adds up fast. AMD Threadripper (non-PRO) chips dramatically reduce compile times compared to mainstream desktop processors, and fully populating the memory channels is just as important: 128–256GB of RAM is common for this workload because shader compiles, large editor sessions, and parallel tooling consume a lot.

For ICVFX and virtual production, you're in a different category entirely.

At that level, the GPU becomes even more critical. AMD Threadripper PRO 9000WX series leads for the most demanding VP workflows, and the GPU recommendation moves to NVIDIA RTX PRO 6000 Blackwell 96GB of VRAM for dense UE5 scenes running alongside multi-application workflows that push well past 62GB. This is a platform decision, not just a component choice.

The RAM and storage rules that apply everywhere:
- 32GB minimum for most users, 64GB+ if you're building lighting that runs for hours
- Two-drive setup at minimum: fast NVMe for OS and software, separate NVMe for active project files
- A third storage drive for long-term archival; cheap per GB, keeps your fast drives uncluttered

Is AI just a bubble?

Maybe but if your team is actually using AI to get work done, the more important question is whether you have enough compute to keep up.

In this article Dustin makes the case that AI is becoming a new layer of compute, and the bottleneck is moving from the model itself to the machine running it. That matters a lot once your workflows need privacy, low latency, and serious VRAM.

Read more:

The Shovel is Real: Why AI is the New Layer of Compute GTC was intense, and there is a massive debate raging right now: Is AI a bubble? Is Jensen just a shovel salesman? Or are we looking at the new bedrock?

The Intel Arc Pro B70 is a good example of why GPU buying decisions should start with the workload.

If you need a strong all-around pro GPU for apps like Premiere, After Effects, or Revit, there are better options. But if you need lots of VRAM for AI inference and local model work, the B70’s 32GB of VRAM starts to look very appealing.

Full review:

Intel Arc Pro B70 Review How does Intel's new Arc Pro B70 32GB professional GPU stack up against the competition in media editing, rendering, AI, CAD, and BIM applications?

The Intel Arc Pro B70 is a reminder that not every GPU is built to do the same job.

The B70 brings something interesting to the table: 32GB of VRAM at a price that makes multi-GPU AI workstations more realistic. That matters if your team is trying to run larger models locally or reduce cloud dependency.

In traditional creative and engineering apps, the B70 is more mixed. It’s fine in some workflows, competitive in others, and not the right choice if you need the strongest all-around GPU for Premiere, After Effects, or Revit. But if your priority is AI inference and memory capacity, that changes the conversation.

That’s the real benefit: this card gives teams another way to build for local AI without jumping straight to the most expensive options.

Full review:

Intel Arc Pro B70 Review How does Intel's new Arc Pro B70 32GB professional GPU stack up against the competition in media editing, rendering, AI, CAD, and BIM applications?

The AMD Ryzen 9 9950X3D2 Dual Edition is a great example of why “fastest” and “best value” are not always the same thing.

Best for: Photoshop, rendering, game development, and CPU-heavy work that benefits from every extra bit of speed.

Not always best for: Premiere users who may get more from Intel’s QuickSync advantage and lower-cost platforms.

That’s why workstation specs should always start with the workflow, not the headline number.

Full review: https://hubs.ly/Q04hYLMF0

Creative work is moving off the desk and sometimes off the cloud.

Joey Daoud (-land on YouTube) had a great conversation with our Founder, Jon Bach, at NAB Show 2026 about the big shifts shaping the PC hardware industry right now: on-prem rack workstations, memory shortages driven by AI demand, and why cloud is often the best place to start before moving to local hardware.

If you’re thinking about how AI, remote workflows, and creator hardware fit together in 2026, this is a great place to start.
Link to the video in the comments. 👇

Being in the right room matters. But who you bring with you? That's everything.

Amy DeLouise, Founder of , sits down with our very own Nicki Sun, Head of Media Production at Puget Systems, on The Puget Experience Stage at 11am for a real conversation about sponsorship, community, and why making space for women in tech and media production is just plain good for the industry.

We're proud to be a sponsor, and even prouder to have Nicki leading that charge for us.